FTL should be cool with runtime functions throwing exceptions

[Filip Pizlo]

This requires an exception throw off-ramp from the FTL.  It's pretty easy.

[Filip Pizlo]

Created attachment 221213 [details]
the patch

[Geoffrey Garen]

Comment on attachment 221213 [details]
the patch

r=me

Would be nice to have a test case for the rare case where an FTL function does not handle any exceptions.

[Filip Pizlo]

(In reply to comment #2)
> (From update of attachment 221213 [details])
> r=me
> 
> Would be nice to have a test case for the rare case where an FTL function does not handle any exceptions.

This gives me the excuse to write this test:

function foo() { }

noInline(foo);

for (var i = 0; i < 100000; ++i) {
    var result = foo();
    if (result !== void 0)
        throw "You broke JSC so hard that even the empty function doesn't work: " + result;
}

[Filip Pizlo]

Landed in http://trac.webkit.org/changeset/162035

[Geoffrey Garen]

Nice!